An Improved 2D Colonic Polyp Segmentation Framework Based on Gradient Vector Flow Deformable Model

Computed Tomography Colonography has been proved to be a valid technique for detecting and screening colorectal cancers. In this paper, we present a framework for colonic polyp detection and segmentation. Firstly, we propose to use four different geometric features for colonic polyp detection, which include shape index, curvedness, sphericity ratio and the absolute value of inner product of maximum principal curvature and gradient vector flow. Then, we use the bias-corrected fuzzy c-mean algorithm and gradient vector flow based deformable model for colonic polyp segmentation. Finally, we measure the overlap between the manual segmentation and the algorithm segmentation to test the accuracy of our frame work. The quantitative experiment results have shown that the average overlap is 85.17% ± 3.67%.